Wednesday, August 20, 2008

Analyzing Central Tendancy Statistics

Analyzing Central Tendancy Statistics

A short sample survey includes merely five questions relating to product use, frequency of use, most recent purchase, future purchase intention and a limited set of purchase intention reasons. Therefore the scope of the study is strictly limited.

Questions requiring consideration:

1. What is the overall purpose of the study?
2. What would be an acceptable sample size?
3. What level of standard error is acceptable?
4. What will the cost of the study be?

Random sample survey method would satisfy time and cost factors which might assemble quite an accurate picture of consumer sentiment in fact described as possibly more accurate than conducting a complete census due to errors inherent in inspection processes.

Issues which effect method of sampling include:

-size and cost limitations
-accessibility of consumer population
-desired accuracy

A simple random sample is easy to select however may not completely represent a particular consumer profile. A systematic sample of known and catalogued prior consumers selected by every fourth or fifth name might be more relevant as their prior consumption is known with the same disadvantages of a random sample. A cluster or judgements based sample could be used on known regular or loyal customers but would take more time to compile. (Evans & Lindsay, Management and Control of Quality Fourth Edition, pp. 524-532)

As variation in responses are likely to be equally valued as in yes or no responses or spread over three choices as in the multiple choice responses several different questions utilize measurement of the mean response and might make sense in terms of a simple range of data, and measurement of the distance of each response from a central tendancy or standard deviation. Therefore mean or average reponses would be useful for yes or no type questions such as #1 or #4. For example questions #1 or #4 do not require mode ranges but merely average response either yes or no. Analysis of reponses to #2, #3 and #5 would require a combination of both median and mode analysis to establish average mean out of the median responses range and mode or most frequent response for more elaborate discussion of the topics. For example, in questions #2, #3 and #5 we could determine the average answer and the most frequent purchasing periods as well as performance indicators all out of median and mode.

Therefore one would seek not to exclusively rely upon one single method of central tendancy however using them all at various purposes as is their intention to represent data to meet the needs of the survey.

No comments: